💡Add `error.file` property on RumErrorEvent to identify the source file of errors

[kazukitaninaka]

Description

I would like to propose adding a file property to the error object in RumErrorEvent. This property would contain the URL of the file where the error originated, extracted from the first frame of the stack trace.

Use Case

When analyzing errors in Datadog RUM, it's often important to distinguish between:

• First-party errors: Errors originating from our own application code
• Third-party errors: Errors from external scripts (analytics, ads, browser extensions, etc.)

Currently, to determine the error source, users must:

• Parse the error.stack string manually
• Extract the URL from the first stack frame
• Implement custom filtering logic in beforeSend

Having a dedicated error.file field would provide a simple way to filter out third party errors that you might not want to investigate

[kazukitaninaka]

DataDog/rum-events-format#331
#4055

[BenoitZugmeyer]

Hello,

Thank you for your contribution! Are you using the error tracking product? Because error tracking already extracts the first stack frame file and you should be able to query errors using it.

[kazukitaninaka]

@BenoitZugmeyer
Thanks for your response. What I am trying to do is to create an anomaly detection monitor based on number of errors collected via RUM.
I think error tracking monitor can only be set based on errors with the same issue id, but I would like to make a monitor that considers any kind of errors.

[BenoitZugmeyer]

I asked the team, and it turns out that if you upload your JS source maps, the @error.file property will be defined in RUM events. See https://docs.datadoghq.com/real_user_monitoring/guide/upload-javascript-source-maps. Let me know if it works for you.

For a bit of context, I'd like to avoid adding more things related to stack trace parsing in the SDK, because I hope this can be done in the backend in the future.